Bioscience Reports, Vol. 20, No. 6, 2000
MINI REVIEW
Mechanisms of Initiation of Membrane Fusion: Role of Lipids Paavo K. J. Kinnunen1,2 and Juha M. Holopainen1 Receiûed August 2, 2000 Main emphasis in studies on the mechanisms of fusion of cellular membranes has been in the roles of various proteins, with far less interest in the properties of lipids. Yet, on a molecular level fusion involves the merging of lipid bilayers. Studies so far have revealed lipids forming inverted non-lamellar phases to be important in controlling membrane fusion. However, the underlying molecular level mechanisms have remained controversial. While this review is focused on presenting one possible mechanism, involving so-called extended lipid conformation, we are also advocating the view, that in order to obtain a more complete understanding of this process it is necessary to merge the relevant physicochemical properties of lipids with the models describing the specific functions of proteins. To this end, taking into account the central importance of fusion in a wide range of cellular processes, we may anticipate its control to open novel possibilities also for therapeutic intervention. KEY WORDS: Membrane fusion; lipid conformational dynamics; lipid packing; extended lipid conformation; fluorescence spectroscopy; pyrene. ABBREVIATIONS: bisPDPC, 1,2-bis[(pyren-1)-yl]decanoyl-sn-glycero-3-phosphocholine; DOPA, 1,2-dioleoyl-sn-glycero-3-phosphatidic acid; DOPC, 1,2-dioleoyl-sn-glycero-3-phosphocholine; DMPC, 1,2-dimyristoyl-sn-glycero-3-phosphocholine; DPH, 1,6-diphenyl1,3,5-hexatriene; HII , inverted hexagonal phase; Ie , pyrene excimer fluorescence emission intensity; Im , pyrene monomer fluorescence emission intensity; IMI, inverted micellar intermediates; Lα , fluid lamellar phase; Lε , frustrated lamellar phase; LDL, low density lipoprotein; LUVs, large unilamellar vesicles; Me2C, divalent metal cation; NBD, 7-nitro-2,1,3benzoxadiazol-4-yl; P, polarization; PA, phosphatidic acid; PC, phosphatidylcholine; PDPTPC, 1-[(pyren-1)-yl]decanoyl-2-[(pyren-1)-yl]tetra-decanoyl-sn-glycero-3-phosphocholine; PE, phosphatidylethanol-amine; PEG, poly(ethyleneglycol); PPDPC, 1-palmitoyl2[(pyren)-1-yl]decanoyl-sn-glycero-3-phosphocholine; sn, stereochemical notation; SUVs, small unilamellar vesicles; X, mole fraction of the indicated lipid.
FUSION OF BIOMEMBRANES: AN OVERVIEW Membrane fusion is centrally involved in a large number of cellular processes, such as membrane recycling, protein trafficking within the cell, exocytosis, fertilization, 1
Helsinki Biophysics and Biomembrane Group, Department of Medical Chemistry, Institute of Biomedicine, P.O. Box 8 (Siltavuorenpenger 10A), University of Helsinki, Finland. 2 To whom correspondence should be addressed at; Institute of Biomedicine, Department of Medical Chemistry, P.O. Box 8 (Siltavuorenpenger 10A), FIN-00014 University of Helsinki, Finland. E-mail:
[email protected] 465 0144-8463兾00兾1200-0465$18.00兾0 2000 Plenum Publishing Corporation
466
Kinnunen and Holopainen
and enveloped virus infection. Fusion involved in the synaptic transmission of impulses in neurons (Jahn and Su¨ dhoff, 1999) now represents probably the most studied and perhaps the best understood example of biomembrane fusion. In the pursuit for an understanding of the machineries controlling the above cellular processes proteins have so far received the main interest (e.g., Hoekstra, 1990; Zaks and Creutz, 1990; White, 1992; So¨ llner et al., 1993; Jahn and Su¨ dhoff, 1999), with less attention to the possible involvement of lipids. Although several proteins are important in initiating fusion this process can be regarded as a lipid level event culminating in the merging of the bilayers. To this end, no protein has been found to account for the Ca2C sensitive step in the fusion of synaptic vesicles, thus readily suggesting a possible role for acidic phospholipids. A very different example for the importance of fusion is provided by the metabolism and pathophysiology of plasma low density lipoprotein, LDL (Kruth, 1997). Although a major fraction of LDL is metabolized in the cells by the receptor mediated pathway, a significant amount is also taken up by non-receptor mediated mechanisms, involving fusion between the plasma membrane of e.g., macrophages and the lipid monolayer surrounding the LDL particle. Similar mechanisms could be involved in the lysosomal processing of LDL (Lauraeus et al., 1998). Fusion of LDL particles in the arterial initima is currently considered to be the major process in the development of atherosclerotic lesions (Guyton and Klemb, 1989). The development of cationic liposomes as vehicles for efficient delivery of DNA into cells, transfection has become one of the key areas in molecular medicine (Felgner and Ringold, 1989). Also this process involves fusion of the liposome-DNA complexes with cellular membranes. Obviously, deciphering the underlying detailed molecular level mechanisms can be expected to yield novel therapeutic means for a variety of disorders. Membrane fusion is a highly localized event, progressing as a sequence of distinct and specific steps. Prior to fusion the two membranes must first come into close contact, which requires that short-range repulsive undulation, electrostatic and兾or hydration forces must be overcome as these factors at atomic distances represent a large energy barrier, prohibiting fusion (Rand and Parsegian, 1986, 1989; Chernomordik et al., 1987). How this repulsive barrier is overcome in biological membranes has remained elusive and represents one of the fundamental questions in understanding the mechanisms controlling fusion in living cells. Two different views to account for this problem have been proposed and are called either the protein or the lipid theory of membrane fusion. In the former (Lindau and Almers, 1995), the fusion pore is described as a toroid formed of the involved proteins and spanning the two adjacent membranes, forming an oligomeric ring structure. In this model lipids would only be directed into the fusion site subsequent to the formation of the protein pore. Moreover, the proposed involvement of lipidic non-lamellar fusion intermediates would be governed by proteins. The lipid theory of membrane fusion assigns a greater role for lipids in forming the fusion intermediate, limiting the role of proteins to the reduction of the free energy barrier inhibiting membrane fusion. Subsequently to the contact formation the membranes adhere. This step is essential yet not sufficient for fusion (Zimmerberg et al., 1993) and may also be reversible. Partial dehydration of the adhering surfaces precedes the formation of an intermediate state and ‘‘hemifusion’’ which allows for lipid mixing between vesicles
Lipid Conformational Dynamics and Bilayer Fusion
467
in the absence of coalescence of the aqueous contents (Zimmerberg et al., 1993; Katsaras et al., 1993; Arnold, 1995). Finally, a fusion pore opens into the septum separating the internal aqueous cavities of the two merging membrane compartments, allowing their mixing (Zimmerberg et al., 1993).
LESSONS FROM LIPOSOMES Because of the fact that at some stage in the course of the fusion process merging of the lipid bilayers must take place, liposomes continue to provide a good model to study the different stages of the mechanism and machinery of fusion. The properties of lipids have been shown to be of importance. For example, Chernomordik et al. (1997) reported that depending on their 3-D molecular shape different lipids either inhibited or promoted influenza virus fusion, and also showed that hemagglutinin initialized membrane fusion by promoting the formation of a lipidintermediate. Corver et al. (1995) found that the fusion of Semliki Forest virus with cholesterol containing liposomes required 1–2 mol.% of sphingolipid to be present in the membrane, lending support for a specific role for these lipids. A recent review on the role of non-bilayer forming lipids in biological fusion (Chernomordik, 1996) concludes that the lipid composition affects the fusion ‘‘downstream’’ to the activation of the fusion proteins and ‘‘upstream’’ to the fusion pore formation. The effects of lipids are strongly correlated with their geometrical shapes. In this context it is relevant to emphasize that it is the effectiûe molecular shape which matters. Accordingly, characterization of the properties of a lipid in this respect requires knowledge not only about its chemical structure, determining its van der Waals volume but also information on its hydration shell, conformation, conformational dynamics, and intermolecular interactions, involving those due to charges. In other words, the effective geometry of a lipid can be considered as a soft and diffuse surface determined by (i) the amplitudes and frequencies of the thermal motion of the molecule, (ii) fields due to its charges and dipoles, (iii) interactions with its neighboring lipids (e.g., hydrogen bondings), and (iv) the hydration shell of its headgroup (Kinnunen, 1996b). The above is completed by information on the distribution of membrane free volume Vf which represents the difference between the effective and the van der Waals volumes per molecule (Bondi, 1954; Turnbull and Cohen, 1970). For a phospholipid bilayer Vf arises from short-lived, mobile structural defects due to trans-gauche isomerization of the lipid acyl chains created because of packing constraints as well as by thermal motion (Xiang, 1993). Vf and its distribution further depend on factors such as hydration (Lehtonen and Kinnunen, 1994). Micelle forming lipids with positive spontaneous curvature such as lysophospholipids having a large headgroup relative to the hydrophobic part inhibit fusion (Chernomordik et al., 1995). Lipids which induce fusion should be cone shaped with the projected headgroup area occupied in the membrane plane being less than that occupied by the hydrocarbon chains, in general corresponding to a lipid structure with a small or weakly hydrated headgroup. Unsaturated 1,2-sn-diacylglycerols represent an example of the former and phosphatidylethanolamines of the latter
468
Kinnunen and Holopainen
category. Further, short chain saturated and long-chain unsaturated 1,2-sn-diacylglycerols increase the fusion rate whereas long-chain saturated species do not (Sa´ nchez-Migallo´ n et al., 1995). In cells 1,2-sn-diacylglycerols are produced by phospholipase C. Given the important role of this enzyme in cellular signaling cascades understanding of the membrane level physicochemical consequences of the non-equilibrium state generated by product accumulation is of great interest (Nieva et al., 1989, 1993, 1995). Another example of fusogenic lipids is provided by the negatively charged species (phosphatidylserine, phosphatidic acid, phosphatidylglycerol, and cardiolipin) which upon binding of divalent cations (most commonly Ca2C and Mg2C ) can trigger fusion. This is likely to result from dehydration and charge neutralization of their headgroups changing their effective molecular geometry. Destabilization of the bilayer structure must precede fusion. Most ambiguity in fusion mechanisms concerns the exact nature of this transient intermediate state and the involved molecular arrangements. As cone shaped and HII phase forming lipids promote membrane fusion, it has been postulated that fusion and lamellar →HII phase transition proceed via common intermediate structures (Siegel, 1986; Verkleij et al., 1979; Hoekstra and Martin, 1982; Ellens et al., 1984, 1986a, b, 1989; Bentz et al., 1985; Walter et al., 1994; Kinnunen 1992, 1996b; Siegel and Epand, 1997). In other words, compounds promoting fusion should either form or promote the formation of the HII phase. Likewise, conditions promoting fusion should also favor the formation of the HII phase. Intervesicle contacts are required for liposomes to undergo lamellar→HII phase transition (Ellens et al., 1986). The rate of fusion of PEs is highest for bilayers in the poorly understood intermediate phase (Lε ) existing between the Lα and HII phases (Ellens et al., 1989). In this intermediate phase thermal motion increases the effective size of the hydrocarbon region and causes negative spontaneous curvature (Tate et al., 1991). In general, the lamellar →HII phase transition is qualitatively easily comprehended in terms of changes in the effective molecular shapes and can be promoted by decreasing the effective headgroup size by dehydration or deionization. Accordingly, similarly to fusion the promotion of the HII phase formation by acidic lipids in the presence of divalent cations, most prominently Ca2C, is likely to reflect both of these processes. The formation of the HII phase is additionally enhanced by factors which influence the hydrocarbon region of the lipids, i.e., decrease in pressure, increase in temperature, acyl chain unsaturation, increase in acyl chain length, and the presence of proper lipophilic additives, surfactants, solvents, metabolites, or chaotropic agents, such as KSCN and urea (Yeagle and Sen, 1986). The effect of the latter is readily compatible with the suggestion that the affinity of water to the phosphatidylethanolamine (PE) headgroup should decrease upon entering the HII phase (Seddon et al., 1983). Accordingly, we may envisage the effect of chaotropic substances to be due to a diminished affinity of individual water molecules to the bulk aqueous phase, as this allows for PE to maintain its hydration shell at higher temperatures, counteracting the thermal increase in the trans → gauche isomerization and the corresponding increase in the effective volume of the acyl chain region (Mantsch et al., 1981). To this end, the sensitivity of the HII phase for hydration and water activity is of great interest in the light of the
Lipid Conformational Dynamics and Bilayer Fusion
469
increasing recent recognition of the importance of osmotic forces in the regulation of cellular functions. Three models for the intermediate structures linking fusion and the lamellar→HII phase transition have been forwarded. The first one was based on the formation of inverted micellar intermediates, IMIs (Verkleij et al., 1979, Siegel, 1984). This view was supported by the appearance of intramembrane lipidic particles IMIs under conditions where fusion or the lamellar →HII phase transition occur. In addition to electron microscopy, evidence supporting the involvement of IMIs was also derived from 31P-NMR which reports an isotropic signal under conditions promoting fusion as well as in the so-called isotropic lipid phases present below the actual onset of the lamellar→HII transition. Yet, direct causality was questioned (Allen et al., 1990; Verkleij, 1984; Bearer et al., 1982; Papahadjopoulos et al., 1990) and evidence revealing the lack of involvement of IMIs in fusion has been forwarded (Caffrey, 1985; Siegel et al., 1994; Hui et al., 1983; Allen et al., 1990; Verkleij, 1984; Bearer et al., 1982; Papahadjopoulos et al., 1990). The isotropic 31P-NMR signal may also arise in lipids retaining the bilayer configuration (Thayer and Kohler, 1981). To conclude, it is now rather generally accepted that IMI are not required for fusion (Papahadjopoulos et al., 1990; Siegel, 1993). The second mechanism is based on the formation of semitoroidal structures, ‘‘stalks’’ (Markin et al., 1984), estimated to have diameters of >4 nm and lifetimes
